Sagemaker Studio notebook - no module named 'tensorflow' when chosen image type "Tensorflow 2.6 Python 3.8 GPU optimized"

Question

Hi,

I have created a simple notebook on SageMaker Studio using the Image "Tensorflow 2.6 Python 3.8 GPU optimized". But when I try to run simple statement viz. "import tensorflow", I am getting the error "no module named 'tensorflow'".

I tried to install 'tensorflow' package using pip from the terminal attached to the image. But it shows the message "requirement already satisfied".

Am I missing anything here? Please help.

Thanks in advance, 
Praveen

Accepted Answer

From the question, I understand that you are trying to use a TensorFlow 2.6 Python 3.8 kernel in SageMaker Studio, but you are unable to import tensorflow.

The service team are aware of this issue and are actively working on a fix.

Mitigation Option

A)  If your use case is version flexible, version other than 2.6 should work.
 
B) If not, you can try the following as workaround

1. Open a notebook using a Tensorflow 2.6 Python 3.8 kernel
2. Execute the following line in a notebook cell:
!sed -i 's|^ *"python",|  "/usr/local/bin/python",|g' /usr/local/share/jupyter/kernels/python3/kernel.json
3. Stop the kernel
4. Re-attach the kernel to your notebook.

Hope it helps!

Sagemaker Studio notebook - no module named 'tensorflow' when chosen image type "Tensorflow 2.6 Python 3.8 GPU optimized"

Relevant content